ISMIR2004 Audio Description Contest - Rhythm Classification

Purpose
The goal was to compare algorithms for automatic classification of 8 rhythm classes (Samba, Slow Waltz, Viennese Waltz, Tango, Cha Cha, Rumba, Jive, Quickstep) from audio data.

Participants
A single algorithm has been submitted to the rhythm classification contest, it was submitted by Thomas Lidy and Andreas Rauber from the Vienna University of Technology and Andreas Pesenhofer from eCommerce Competence Center in Vienna, Austria.

Data
Participants were given a list of 488 training instances in order to train their systems (30-s instances corresponding to the 8 rhythmic classes, real audio decompressed to .wav format, 44100 Hz, 16 bits, mono). (See instructions given to participants here.) These files could be downloaded from http://www.ballroomdancers.com/Music/style.asp
The list of training instances is here.
A test dataset of 210 instances (same format, coming from the same source) was used to evaluate the submitted algorithm.
The list of test instances is here.

Evaluation
The evaluation metric was the percentage of correctly classified instances.

Results
The algorithm accuracy was of 82%.